Institutions, Technology, and Trade

2 downloads 116 Views 972KB Size Report
An earlier draft of this paper circulated under the title "Tariffs, Trains, and Trade: The Importance of Institutions versus Technology in the Expansion of Markets".

NBER WORKING PAPER SERIES

INSTITUTIONS, TECHNOLOGY, AND TRADE Wolfgang Keller Carol H. Shiue Working Paper 13913 http://www.nber.org/papers/w13913

NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 April 2008

An earlier draft of this paper circulated under the title "Tariffs, Trains, and Trade: The Importance of Institutions versus Technology in the Expansion of Markets". We thank Lee Alston, Jörg Baten, Brad DeLong, Price Fishback, Rainer Fremdling, Murat Iyigun, Michael Kopsidis, Larry Neal, Nathan Nunn, Florian Ploeckl, Steve Redding, Kevin O.Rourke, John Sutton, Je¤ Williamson, Randy Wright, and the audiences at Berkeley, Colorado, Harvard, Yale, the 2006 NBER/CEPR conference in Lund, the 2007 NBER Spring ITI meeting, the 2007 North American Cliometrics meeting, the 2008 NBER-Philly Fed Macroeconomics in Time and Space conference, and the 2008 CEPR ERWIT conference for useful comments, and Michael Kopsidis as well for providing us with data. NSF support under grant number SES 0453040 is gratefully acknowledged. The views expressed herein are those of the author(s) and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications. © 2008 by Wolfgang Keller and Carol H. Shiue. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.

Institutions, Technology, and Trade Wolfgang Keller and Carol H. Shiue NBER Working Paper No. 13913 April 2008, Revised December 2008 JEL No. F1,F3,N10,O24,O3 ABSTRACT We study the importance of technology and institutions in determining the size of markets in five different countries and fifteen different German states. The setting of 19th century Europe presents a unique opportunity to address this issue, since it witnessed fundamental change in both dimensions. At the beginning of the century, numerous customs borders, separate currencies with different monetary systems, and poor transportation facilities were major obstacles that held back trade. Important institutional change, through the Zollverein customs treaties and currency unification, and major technological innovations in the steam train all had a role in increasing market size as measured in terms of the spatial dispersion of grain prices across 68 markets. However, we find that the impact of steam trains is substantially larger than the effects from customs liberalizations and currency agreements in increasing market size, where correcting for the potential endogeneity in institutional and technological changes are crucial for this result. We also find that a state's institutions influence the rate of adoption of steam trains, thereby identifying an important indirect effect from institutions on economic performance. The institutional and technological changes account for almost all of the decline in price gaps over this period.

Wolfgang Keller Department of Economics University of Colorado at Boulder Boulder, CO 80309-0256 and NBER [email protected] Carol H. Shiue University of Colorado at Boulder Department of Economics Boulder, CO 80309 and NBER [email protected]

1

Introduction

The growth of Europe, and indeed of the world, has been linked to the scope of market transactions. In fact, it has been said that economic development is the spread of markets.1 As the market size increases, trade raises economic welfare due to the e¢ ciency gains of greater specialization, the division of labor, and potential additional gains through scale economies. Moreover, the welfare gains from increases in market size can be compounded manifold if they usher in an era of higher sustained per-capita growth for an economy. Institutions that reduce transactions costs are one obvious factor that determines the size of the market, yet little is known about the importance of institutions relative to other means by which market expansion occurs.2 This paper analyzes these issues by examining the spatial dispersion of grain prices in 68 European markets. These markets are located in …ve di¤erent countries and …fteen di¤erent German states, including Prussia. The area corresponds approximately to the location of today’s Austria, Belgium, Czech Republic, France, Germany, Italy, the Netherlands, Poland, and Switzerland (See Figure 1). The goal of this paper is to assess the contributions of the key institutional and technological innovations to trade in 19th century Europe, when the scope of markets was increasing at a speed faster than at any point in the past. The …rst question we ask, and answer, is how much of an impact did institutional agreements have on market size compared to that of a key transport technology of the 19th century, the steam train. Transactions costs to trade were high in the late 18th century. The steam train dramatically altered the transportation system in which overland transport was often 1

Braudel (1992, 225). We borrow North’s (1990, 3) notion of institutions; they are "the rules of the game in a society or, more formally, are the humanly devised constraints that shape human interaction". 2

1

achieved by horsedrawn carriages. There were about 1,800 customs borders in Germany alone at this time.3 Moreover, a vast number of di¤erent currencies existed, which had the e¤ect of reducing market size even further. Fundamental change arrived in the 19th century. The main mechanism bringing down customs borders was an institution called the Zollverein, the "classic example of a customs union".4 Starting in the year 1828, the Zollverein treaties successively liberalized trade among some thirty-…ve member states that would later become Germany. The …rst half of the 19th century witnessed also the creation of the …rst major monetary agreements in Europe. The institutional framework provided by these laws governing commodity and foreign exchange transactions was a clear break from centuries of relatively chaotic conditions, and so one would expect their impact to be large. We show, however, that accounting for the endogeneity of these changes is critical for obtaining proper estimates of the impact of institutions on the size of the market. Secondly, we consider whether the timing of the adoption of technology could be an indirect outcome of the di¤erent institutions that prevailed in 19th century Europe. Over the last several years the literature has emphasized fundamental institutions in bringing about a major impact on economic outcomes (Hall and Jones 1999, Acemoglu, Johnson, and Robinson 2001, Rodrik, Subramaniam and Trebbi 2002). In this line of research, institutions may have a direct e¤ect on economic outcomes, as well as an indirect e¤ect, because they in‡uence the form and rate of technological progress and they shape the incentives of agents (Helpman 2008). We shed new light on this question by examining whether the impact of steam trains on market size is a¤ected either by proximate or by more fundamental institutional factors. As a proximate determinant, we ask whether trains have a stronger impact when they are run by private agents, as opposed 3 4

Henderson (1959, 21). Viner (1950, 97).

2

to the state.5 We also examine whether steam trains had a stronger impact on market size in states that abolished serfdom relatively early, hypothesizing that early abolition of serfdom is a general sign of institutional quality.6 Our setting has a number of advantages. First, by covering the entire 19th century, we can observe many episodes of change. Moreover, the economies are going through both institutional and technological change, so that the impacts for the same set of economies can be compared. Second, the economies of 19th century Europe also proceeded at di¤erent speeds in the implementation of customs, currency, and transportation changes. This is crucial since one can look for corresponding e¤ects on market size at di¤erent times across economies. Third, these economies also had di¤erent institutional fundamentals, and this allows us to make some headway towards separating the direct and indirect e¤ects of institutions on market size. We …nd that both institutional change through currency agreements and customs liberalizations and technological change through the adoption of steam trains were important in increasing the size of the market in 19th century Europe, accounting for almost all of the decline in price gaps. However, the impact of steam trains is found to be larger than the customs and currency e¤ects. We also …nd support for both proximate and fundamental institutions a¤ecting technological change, thereby providing some initial evidence for indirect e¤ects of institutions via their impact on the rate technological change. Transportation technology improvements have often been seen as a crucial factor for reshaping the geography of trade. The arrival of more e¢ cient transatlantic shipping led to important 5 This question is often emphasized in the literature, with the general presumption that private railways would be more e¢ cient; for the case of Germany, see, e.g., Fremdling (1975, 109-132). 6 Acemoglu, Cantoni, Johnson, and Robinson (2008) study the impact of the French Revolution, showing that the existence of serfdom, as one of the core feudal institutions, is a good indicator for institutional quality, and negatively correlated to economic outcomes.

3

changes in the late 19th century (Harley 1988, O’Rourke and Williamson 1999), while container shipping and air freight may have increased o¤shoring during the late 20th century (Hummels 2007). While many writers would agree that steam railways revolutionized transport technology during the 19th century, not all earlier studies bear out the large impact of rails (Fogel 1964, Fishlow 1965, and Williamson 1980).7 One advantage of our approach to transport technology is that it employs economy pair-speci…c information on the establishment of rail connections, and moreover, this includes crucial geographic factors a¤ecting steam train availability.8 This analysis is unique in comparing causal e¤ects from customs liberalization, currency agreements and transportation innovations using economy-speci…c information on all three mechanisms. Even if there is general agreement that declining transport costs, both now and in the 19th century, have a bearing on world trade, the recent emphasis has shifted to non-transport cost factors, especially payment frictions (Alesina and Barro 2002, Eichengreen and Irwin 1995, Flandreau and Maurel 2001, Frankel and Rose 2002, and Rose 2000).9 Several authors have also simultaneously examined, as do we, the impact of customs liberalizations.10 In this respect, our work is closely related to Estevadeordal, Frantz, and Taylor (2003) and Lopez-Cordova and Meissner (2003), who estimate that bilateral trade in the late 19th century was substantially higher if both partners were signatories of a monetary agreement, the gold standard, than if 7

Cameron and Neal (2003, 199); they go on to write "[Railroads] were both the symbols and the instruments of industrialization. Before the railways inadequate transportation facilities constituted a major obstacle to industrialization in both continental Europe and the United States." 8 Estevadeordal, Frantz, and Taylor’s (2003) main information on transport cost is Isserlis’ (1938) freight cost index, which is based on British rates. For a global sample, this introduces measurement error; as noted in Isserlis, e.g., British ships carried less than 10% of trade between two foreign ports at the time (p.139). Jacks and Pendakur (2008) suggest to combine the maritime trade cost information with overland rates, noting that O’Rourke and Williamson (1994) show that much of the decrease in price di¤erentials between the UK and US after 1850 was due to lower transport costs between the Midwest and the East Coast of the US. 9 Glick and Taylor (2008) document the disruptive e¤ects of wars on trade. See also Jacks (2005, 2006) who studies a broad range of transportation and non-transportation factors. 10 On customs liberalizations, see also Baier and Bergstrand (2007), Rose (2004), and Subramanian and Wei (2006).

4

they were o¤ the gold standard. Our study di¤ers from theirs in that we consider a period starting 70 years earlier. We also conduct an instrumental variable analysis that is compelling, we believe, in picking up on key historical facts to establish that the timing of institutional change in 19th century Europe was systematically related to the economies’speci…c costs and bene…ts. The results are substantively di¤erent from what comes out of treating institutional change as exogenous.11 The remainder of the paper is as follows. Section 2 highlights the convergence of prices in 19th century Europe, and presents preliminary evidence on the importance of institutional and technological change in this. We also provide some historical background for customs liberalization, currency agreements, and steam train adoption during this period. Our data is decribed in section 3. The empirical results on their importance for increases in market size are presented in section 4, where we also report evidence for the impact of institutions on the rate of technological change. The concluding section 5 summarizes our results and suggests directions for future research.

2

Price Convergence in 19th Century Europe: the Zollverein, Currency Agreements, and Steam Trains

Consider the European cities Berlin, Brussels, and Munster. The …rst was the capital of Prussia, and later, of the German Reich. The second is the capital of Belgium (founded in 1830), and 11

Estevadeordal, Frantz, and Taylor (2003) …nd instrumental variable (IV) estimates that are similar to OLS estimates, while Lopez-Cordova and Meissner’s (2003) …nd no evidence for endogeneity. Ritschl and Wolf’s (2003) focus is on showing that a high propensity to trade leads to currency agreements rather than estimating the causal impact of them on trade. For the later part of the 20th century, Barro and Tenreyo’s (2004) IV analysis shows that the strong e¤ect of currency unions on trade found in Rose (2000) and others is not subject to endogeneity bias.

5

Munster was the capital of the Prussian province of Westfalia. Figure 1 shows the location of all three cities on a map. Between Berlin and Munster, customs borders were removed in the year 1831, while between Berlin and Brussels, customs borders remained. Does this customs liberalization explain why between 1830 and 1855, the Berlin-Munster relative price gap for wheat fell by 0.12, while for Berlin-Brussels it fell only by about 0.02? A priori, it is plausible that customs liberalization did indeed play a role. However, the period 1830-55 witnessed not only customs liberalization, but also, in the year 1848, the arrival of a steam train connection between Berlin and Munster, while a train connection between Berlin and Brussels was not available until 1859. Hence, it would be hard to argue that the di¤erence in price gap reduction of 0.10 (0.12 minus 0.02) is due to customs liberalization alone. There are also the German cities of Nurnberg, and Parchim. The former is a major city in Bavaria, while Parchim was a smaller town in the state of Mecklenburg-Schwerin, in Germany’s far North (see Figure 1). Between Berlin and Nurnberg, it was possible to transport wheat on trains from the year 1851 on, while between Berlin and Parchim, such transport was possible only from the year 1880 on. Does this di¤erence in the transport options explain why between 1830 and 1855, the Berlin-Nurnberg relative price gap for wheat fell by 0.18, while for BerlinParchim it fell only by about 0.01? Trains may have brought down the Berlin-Nurnberg price gap faster than for Berlin-Parchim. However, Berlin and Nurnberg also became members of the Zollverein in 1834, while Parchim joined it only in 1867. Moreover, it is surely no accident that trains connected the relatively important cities of Berlin and Nurnberg three decades before it was possible to transport grain by steam train from Berlin to the less important town of Parchim. The marginal contribution from trains on relative price gaps may therefore be di¤erent than what these …gures indicate.

6

As a …rst cut, then, the evidence is consistent with customs liberalization, train transport, and also currency agreements each expanding market size. At the same time, it is also instructive to review what is known about the perceptions of those who lived in these areas at the time. It was recognized early on that internal trade in Germany was hampered by the multiple customs borders. The economist Friedrich List, head of the Union of Merchants (der Deutsche Handels- and Gewerbeverein), expressed this in a petition to the German parliament in the year 1819 as follows: the numerous customs barriers “cripple internal trade and produce the same e¤ect as ligatures which prevent the free circulation of blood. The merchants trading between Hamburg and Austria, or Berlin and Switzerland must traverse ten states, must learn ten customs tari¤s, must pay ten successive transit dues. Anyone who is unfortunate enough as to live on the boundary line between three or four states spends his days among hostile tax-gatherers and customs house o¢ cials. He is a man without country.”12 The customs situation in Germany was also unfavorably compared with that in other major countries, such as France. In the words of List, the situation is “depressing for [German] men who want to act and trade. With envious eyes they look across the Rhine river, where a large nation, from the Canal to the Mediteranean Sea, from the Rhine to the Pyrenees, from the border with the Netherlands to Italy, engages in trade on open rivers and roads without ever encountering a single customs o¢ cial.”13 12 13

The petition is printed in German in von Eisenhart Rothe and Ritthaler (1934, 320-324). Ibid.

7

The support for a removal of customs borders was broad and went beyond merchants, agriculturalists and industrialists.14 For example, Goethe emphasized both the importance of currency agreements as well as customs liberalization. He said that he would look forward to a time when “the German Thaler and Groschen will have the same value throughout the entire country and my luggage may pass unopened through all thirty-six German states.”15 There were also some voices opposing economic liberalization in Germany, especially in the early 1800s. They included political progressives, who would however typically not oppose liberalization per se, but liberalization under the leadership of Prussia, which they considered as politically undesirable.16 These views waned over time, with the increasing recognition that there was no way to German economic liberalization other than under Prussia’s leadership. Also the nobility leading the smaller and mid-sized German states was often hesitant about economic liberalization in Germany, mostly because they feared the possibly accompanying political changes that would result in a loss in their personal power. However, at times the economic imperatives were overwhelming. For example, Ludwig I, the king of Bavaria, strongly supported customs liberalizations with Prussia in the year 1833, because he expected that the economic costs imposed by customs borders would fuel political unrest in the population, thereby leading to a revolution and a loss of his legitimacy (Hahn 1984, 73-75). The following gives a brief overview of how customs and currency agreements as well as steam trains changed the economic landscape in 19th century Central Europe. 14

See Henderson (1959), Hahn (1984). This is not to say that merchants, agriculturalists and industrialists everywhere were in favor of customs liberalization. For example, industrial producers in the South German states of Baden, Bavaria, and Wurttemberg complained about the new competition from Prussian and Saxon products after the 1834 Zollverein treaty (Hahn 1984, 94). 15 Goethe in conversation with Eckermann in the year 1828; see Goethe (1828). 16 Several of the Southern German states had by then adopted constitutions, which Prussia had not.

8

2.1

The Zollverein

The Zollverein was the most important institutional development in the move towards trade liberalization in 19th century continental Europe. Other treaties can be found, but none were as encompassing or long-lived. The main economic impact of the Zollverein treaties was the abolishment of tari¤ barriers among member states, and the implementation of a single external tari¤ for non-members. As of 1815, Germany’s political structure was divided into the thirtynine states of the German Confederation (Deutscher Bund), see Figure 1. The confederation consisted of sovereign states in which joint action depended upon unanimity. Austria was initially the most powerful of the German states, followed by Prussia. Individual states tended to be highly protectionist and the tari¤s that were imposed were complicated. There is no reliable information on enforcement, but it was likely that it was costly especially for the many small states to each monitor its own borders. In the aftermath of debts from a decade of war, and new tari¤s raised by Britain, Russia, Austria, France, and the Netherlands, Prussia sought to negotiate treaties with her neighbors while reforming internal tari¤s. This was particularly pressing because Prussia’s territories were divided into two, an eastern portion consisting of seven provinces, and a western portion that included the Rhineland provinces and the Ruhr area. In the year 1818, the Prussian Customs Union was formed. With few exceptions, internal dues were abolished, and by 1821 only a single tari¤ for the entire Kingdom was levied, while transit dues on goods passing through Prussia were reduced. The importance of the Prussian Customs Union stems from the fact that it served as a model for most of the Zollverein treaties that followed. Enclaves within Prussia were the …rst to develop agreements with Prussia on how its payment of duties were to be treated— with Prussia deciding to treat the enclaves as her own 9

territory rather than as foreign states required to pay import duties. As with all of the following treaties, these were based on the principle that states that adopted the Prussian system of tari¤ received a share of the joint revenue based on population size. Their rights as sovereign states were maintained.17 Hesse-Darmstadt was the …rst territorially separate state to join the Prussian Customs Union in the year 1828, while Hesse-Cassel became the next to join in 1831. The latter was signi…cant because it meant that the East and West Prussian provinces were joined without a customs border for the …rst time. It also meant that British goods could no longer reach Frankfurt and Germany’s south without crossing a Prussian external tari¤ border; see Figure 1. In the year 1834, the Thuringian states, the Kingdom of Saxony, and the 1828 formed South German Customs Union (consisting of Wurttemberg and Bavaria) joined the augmented Prussian Customs Union to become the German Zollverein. At that point the Zollverein had an area of about 163,000 square miles and a population of about 23.5 million people. By stages, other states entered. Three other German states joined the Zollverein between mid-1835 and early 1836: Baden, Hesse-Nassau, and the Free City of Frankfurt. The entry of Baden was signi…cant because it meant that the two separate areas of Bavaria were joined without custom borders. The entry of Frankfurt meant that it was possible to trade manufacturing goods from Frankfurt up the Main River to Northern Bavaria in exchange for grain without paying customs duties. Later on, Brunswick became a member of the Zollverein in 1841, Hanover in 1851, Oldenburg in 1852, and Mecklenburg as well as the Free City of Lübeck in 1867. Two states became members of the Zollverein only after Germany became politically uni…ed in 1871, namely the Free Cities of Bremen and Hamburg in 1888. Thus, the process 17

Throughout, Prussia reserved the right to negotiate with foreign countries such as France, Belgium, and England for itself.

10

of customs union enlargement materialized over a large part of the century (the years 1828 to 1888). Austria-Hungary did not become member of the Zollverein. There are a number of possible reasons for states wanting to join the Zollverein. Market access was certainly a major motive. For any state considering whether or not to join the Zollverein, one advantage was tari¤-free access to the large market of Prussia, which included the leading industrial areas of Germany at this time. In general, however, the states located in Germany’s South joined the Zollverein earlier because not joining implied having to pay hefty tolls in order to reach the Baltic or North Sea coast. This was important, …rst, because it gave access to trade with the emerging industrial powers, in particular England. Moreover, it was the Southern states’main access to sea ports, since the Alps e¤ectively blocked o¤ trade to southern ports. Thus, the Southern German states of Baden, Württemberg and Bavaria had all joined the Zollverein by 1836, whereas the Mecklenburg states, located directly on the Baltic coast, joined only in 1867, and the city states of Hamburg and Bremen, which relied particularly heavily on international trade, joined only in 1888. Fiscal reasons may well have also been part of the calculus, but it is di¢ cult to …nd general patterns. For many of the relatively small states, it was prohibitively costly to establish and enforce tari¤ borders, and they preferred joining the Prussian-led customs union in exchange for a fraction of the joint tari¤ revenue (Dumke 1976, Chapter 1). At the same time, this cannot be the full explanation since there were several highly indebted and small states that joined the Zollverein relatively late. Some of the smaller and mid-sized German states may also have hesitated to join the Zollverein because they preferred more trade protection than the external tari¤ preferred by Prussia provided. However, Prussia’s tari¤s on a range of goods, especially Kolonialwaren such as tobacco, tea, and sugar, were actually higher than the tari¤s of other

11

German states before they joined the Zollverein, so the desire for more protection can hardly be the main reason for not joining the Zollverein.18 Other reasons for joining the Zollverein were idiosyncratic. For example, Hanover joined relatively late in part because it was governed in personal union with England, which had no interest in an all-inclusive Prussian led customs union in the center of Europe. Overall, the key motive for joining the Zollverein was likely market access, and we will return to this question in the empirical analysis below. Since the Zollverein was a customs union, joining it was not identical to a move towards multilateral free trade. Trade diversion was a possible outcome. However, most of trade of the German states at the time was with other German states, and a substantial share of imports were consumption goods that were hardly produced in Europe (such as tobacco, sugar, and spices). Therefore, the trade diversion e¤ect of the Zollverein was rather limited, and thus the basic character of the Zollverein was trade-liberalizing..

2.2

Currency Agreements

In the …rst decades of the 19th century, Germany was replete with coins issued by its many di¤erent states. The diversity was immense, in sharp contrast to the uni…ed monetary conditions in Great Britain and France, for example.19 In the Southern states, the dominating currency was the Gulden, as it was also called in the empire of Austria-Hungary, while in the Northern states the currency was typically called Thaler. Irrespective of the name, each state minted its own currency, and initially currencies did not have legal-tender status outside of a given state. The currencies were linked to silver by the currency unit expressed in equivalent to a certain quantity of silver weighted in Cologne Mark. Comparability of coins even of the same 18 19

See Dumke (1976) for more details. Holtfrerich (1989, 1993).

12

denomination, like Gulden, was di¢ cult because the mints in di¤erent states had di¤erent coinage fees. This meant that the net silver weight of Gulden from di¤erent states would actually di¤er. During the 1820s, the state of Hesse-Nassau for example went as far as to melt down high-silver content coins issued in Bavaria to produce its own low-silver content coins, and pocket the di¤erence (Holtfrerich 1993). The dividing line between full-value specie money and debased coins was therefore ‡uid. The South German states put an end to this through the Munich Coin Treaty of 1837.20 It stipulated that the silver content of the Gulden should be the same (nine-tenth of face value), no matter which state minted it. This e¤ectively meant the …xing of exchange rates among the Southern states’ currencies from this date on. Importantly, Gulden coins minted in any of the Southern states would have legal-tender status in all signatory states. One year later, the Dresden Coin Convention in 1838 e¤ectively led to …xed exchange rates between all Zollverein currencies by requiring that each state was obliged to mint coins according to the common metal-content speci…cations. However, the 1838 Dresden agreement did not give legal tender status to all currencies throughout the Zollverein. This created an important barrier to commercial exchange. The Dresden agreement left the Northern Thaler bloc and the Southern Gulden bloc intact, even though currencies in both blocs were linked to the Cologne Mark at a …xed exchange rate of 1 Thaler = 1.75 Gulden. It was recognized at the time that a generally accepted medium of exchange is important for facilitating trade. In fact, the states agreed on the minting of a common coin worth 2 Thaler or 3.5 Gulden that would have full legal tender status throughout (called the "Vereinsthaler"). In part because its denomination was too large for everyday small-scale business, the coin 20

These Southern states are Bavaria, Baden, Württemberg, Hesse-Nassau, Hesse-Darmstadt, and the Free City of Frankfurt.

13

never played the role for which it was introduced.21 Instead, the Prussian one-Thaler piece was increasingly used for commercial transactions after 1838, and even gained de facto acceptance in the Gulden states of Southern Germany. The need for a generally accepted medium of exchange was remedied only twenty years later, in the Vienna Coin Treaty of 1857, where all Zollverein currencies were given full legal tender status throughout the Zollverein (even retrospectively to those coins minted between 1838 and 1857). The states that remained outside the Vienna currency agreement of 1857 in our sample are Mecklenburg-Schwerin and the Free Cities of Bremen, Hamburg, and Lübeck (Willis 1896). Monetary uni…cation was achieved with political uni…cation of Germany soon after the year 1871. The newly created Reichsmark had full legal tender status in all German states. Also, Germany moved from the silver to the gold standard after the year 1871, in line with the international trend at the time.22

2.3

The Introduction of Steam Trains

European economic growth from the 19th century on also coincided with a series of innovations in transportation.23 These innovations included paved roads, improvements in waterways, railways, in materials such as iron and steel, and later on, steam power, but the rapid increase of railway construction was particularly important. In the 1830s and 1840s British suppliers of locomotives dominated the market, and railway iron exports were an important export for Britain, while countries on the continent started to produce their own railway inputs at a later 21

The signatories expected that by the year 1842, the Vereinsthaler would account for 1.2% of the total coin circulation in Germany. In fact, the Vereinsthaler circulation fell well short of this; Holtfrerich (1993). 22 In our sample, the Netherlands was on the gold standard by 1875, while Belgium and France were on the gold standard by 1880 (Lopez-Cordova and Meissner 2003). 23 A good survey is O’Brien (1983). On the debate concerning the contribution of railways in the United States, see Fogel (1964), Fishlow (1965), and Williamson (1980).

14

stage. The …rst German railway was opened in December 1835. With only 4 miles of tracks, it was a short suburban line located in Bavaria, between Nurnberg and Fürth. The …rst longer route (70 miles) was built in Saxony in 1839, some 5 years after the initial Zollverein treaties came into e¤ect. Thereafter, additional miles of rail were laid down swiftly. By 1847, there were over 2,000 miles of rail in Germany (Henderson 1959, 147), and almost all main railway lines were completed by 1877 (Milward and Saul 1977, 42). Railway building in the …ve sample countries other than Germany, namely Austria-Hungary, Belgium, France, the Netherlands, and Switzerland, proceeded in quite di¤erent ways. In France, railway construction began as early as 1828 with 23 kilometers of track opened, but its pace fell behind that of Germany in part because of resistance to the new technology from owners of other means of transportation. It has been argued that railway building in Germany has been particularly fast because the various politically independent states competed for transport routes through their territitories (Fremdling et al. 1995). At the same time, railway building in Belgium was also very swift. The Belgium railways were designed as a means of international transport from the beginning. This meant that negotiations among di¤erent states were necessary. In 1834, the Belgium Parliament planned for a network that allowed connections to Prussia, France, England, and the sea at Anvers, and later, an extension to Holland (La¤ut 1983). In Switzerland, both the di¢ cult geography as well as the highly federalistic (cantonal) system slowed down railway building. Also in Austria-Hungary, railway building proceeded at a moderate pace; major reasons for that include relatively little interest in the new technology among the empire’s leaders, as well as empty state budgets and lost wars starting around the mid-19th century.

15

How important were railways as a means of transportation for grain? Generally, railways were important for low value-to-weight ratio good such as coal, construction materials, metal goods, and also grain (O’Brien 1983, 1-2). At the same time, the importance of railroads for transporting grain varied greatly. While it was cheaper to transport grain by railroads than by other means of land transport, trains could not compete with transport by ship. In the late 19th century, for example, sending grain from Posen (in East Prussia) to Cologne by train was at least three times as expensive as shipping it to Rotterdam or Antwerp and then up the Rhine river (Köttgen 1890, 64). Consequently, long distance grain trade in the southeast direction, parallel to the major rivers (Elbe, Rhine, and Danube), was hardly ever done by rail. At the same time, transportation of grain on railways was of utmost importance when it connected the drainage areas of the main rivers.24 Grain transportation on railways was also of major signi…cance whenever sea or river transport, even if indirect, was not an option. For example, the great majority of all grain exported from Bavaria to Switzerland in the early 1850s was transported on railways (Seu¤ert 1857, Chapters 5, 6). The attractiveness of transporting grain on railways was not only a¤ected by geographic features. Also the freight rates per ton-kilometer mattered, and while we do not have fully detailed information on this, we know that they di¤ered both across states as well as over time (Hohorst and Fremdling 1979, 64-65). The existence of a train connection therefore does not say everything on the importance of a particular train track for grain trade. The signi…cance of railways for grain trade varied for a number of reasons between di¤erent market pairs, and the estimation will take this into account. 24

For example, the completion of the Köln-Mindener railway in the year 1847 was crucial for transporting the relatively cheap Prussian grain to the emerging industrial areas of the Rhine-Ruhr (Fremdling and Hohorst 1979, 64). At the same time, the availability of paved roads and canals also in‡uenced how important steam railroads were.

16

We now turn to a description of the data.

3

Data

This study employs the price for wheat across markets in Europe to analyze trade and the size of the market. We have compiled a data set consisting of sixty-eight market locations; Table 1 provides an overview. There are 16 markets, or about 24% of the sample, in Austria-Hungary, Belgium, France, the Netherlands, and Switzerland. The remaining 52 wheat price series are for markets located in …fteen di¤erent German states.25 The prices are averages for an entire year, which is appropriate since we are interested in low-frequency changes of price gaps over an entire century. All prices are quoted in terms of Bavarian Gulden per Bavarian Schae¤el (about 223 liter of wheat). To arrive at a comparable set of prices we have converted the many di¤erent quantity and monetary units that were used in 19th century Europe using the conversion rates given in Seu¤ert (1857) as well as in the original sources.26 The overall sample period is 1800 to 1899, but data availability varies greatly across the series. For example, there are all 100 annual price observations for the city of Brugge during the 19th century, while for the market in Wiesbaden, there is only one single observation. Since the goal is to rely on important time-series variation (before-after comparison), it is clear that 25

These German states are the Grand Duchy of Baden, The Kingdom of Bavaria, the Duchy of Brunswick, the Free City of Bremen, the Free City of Frankfurt/Main, the Free City of Hamburg, the Free City of Lübeck, the Kingdom of Hannover, the Electorate of Hesse-Cassel, the Grand Duchy of Hesse-Darmstadt, the Duchy of Hesse-Nassau, the Grand Duchy of Mecklenburg-Schwerin, the Kingdom of Prussia, the Kingdom of Saxony, and the Kingdom of Württemberg. Some of these territories changed their name during the 19th century, for instance the Kingdom of Hannover, which was an Electorate until 1814. All of these territories became part of the German Reich after the year 1871. 26 The rates of in‡ation might have di¤ered across markets, but information on in‡ation rates or exchange rates for all currencies and all years are unfortunately not available. Both the levels as well as di¤erences in in‡ation rates across states have tended to fall over the 19th century, and the currency agreements have likely played a role in this. We include state-pair …xed e¤ects in the analysis to reduce any bias.

17

more weight should be placed on markets where prices are observed for a long time. Table 1 reports the number of observations for each market as well as the year of the earliest price observation during the 19th century. Grain prices in Europe at the time generally increased from the South (the Black Sea area) and East (Eastern Prussia) to the Northwest (Northern Germany, the Canal region, and England). The average percentage price gap between two markets in our sample is about 0.18 over the entire 19th century. For the subset of price gaps for which we have information for the entire century, the average price gap in the …rst decade of the 19th century is about 0.32, while in the last decade of the 19th century it is, with 0.10, less than one third of that. This decline re‡ects that dramatic extent of price convergence over the 19th century. To understand the roles of customs liberalization, currency agreements, and steam railways in price convergence, we coded data on each. For each city-market, we have recorded the year in which it became part of the Zollverein; this year is listed in Table 1.27 Important accession dates are 1834 and 1836, as well as the years 1841 (Brunswick), 1854 (Hanover), 1867 (Mecklenburg and Lübeck), and 1888 (Bremen and Hamburg). Generally, joining the Zollverein meant that barriers for grain trade between any two of its markets would be equal to zero. Unfortunately, there is no comprehensive information on the levels of tari¤s on grain before liberalization. Some available …gures suggest that the duties on wheat may have been on average the equivalent of about 10 percent ad valorem.28 Instead of exploiting the size of the tari¤ change, we rely on the timing of the move towards zero trade barriers through Zollverein 27

Customs liberalizations that did not involve Zollverein accession are discussed in section 4. At the time, mostly speci…c duties were charged, so that the ad valorem duty varied with the price of wheat. In the year 1831, the augmented Prussian customs union charged a speci…c duty equivalent to about 7% for wheat. The ad valorem equivalents for "products of agriculture" before the formation of the Zollverein in 1834 were about 16% in Prussia, 9% in Bavaria and Wurttemberg, 8% in Baden, and 3% in Saxony (Dumke 1976, Tables 3.16, 3.17). 28

18

membership.29 Even though within states tari¤s were generally abolished in the very early 1800s, there could still have been customs borders faced by agents trading within the same state. This is in part because the territory of several states consisted of several non-contiguous parts, such as the Eastern and Western provinces of Prussia, or the Bavarian Palatinate area that was separate from core Bavaria around Nurnberg and Munich. For each market pair in our sample, we have established using historical maps whether a direct trade route would involve passing any customs borders. If the number of customs borders to be crossed is greater than or equal to one, CUijt is coded as 0, otherwise it is 1, for each market pair ij and year t. For any relationship between a German and a non-German market, or between two markets in di¤erent European countries, CUijt is equal to 0 for all years. For example, the customs variable CUijt turns to 1 for the pair Berlin-Nurnberg in 1834 (the year of the Zollverein foundation), it changes from 0 to 1 for Berlin and Parchim in 1867 (with Mecklenburg-Schwerin’s Zollverein accession), and between Berlin and Brussels, CUijt takes on the value 0 throughout the sample period (see Figure 1). Turning to monetary agreements, a major step was for currencies to have full legal tender status in other states. As discussed above, this occurred between the Southern states in 1837 with the Munich Coin Treaty. For all Zollverein currencies, full legal tender status was agreed upon with the Vienna Coin Treaty of 1857. Thus, for example, the variable LTijt for the pair of Munich (in Bavaria) and Stuttgart (Wurttemberg) up to the year 1837 is equal to 0 and 1 afterwards. In contrast, the variable LTijt for the pair Berlin (Prussia) and Stuttgart is 0 up to 29

In a few cases, the time of the Zollverein accession does not coincide with the year in which tari¤s on grain were eliminated. For example, the tari¤s between Bavaria and the augmented Prussian customs union were eliminated in 1829, four years before the initial Zollverein treaty. We focus nevertheless on the Zollverein accession date, because this played the key role in terms of commitment.

19

the year 1857, and 1 afterwards. For relations between a German and a non-German market, LTijt is always 0. Table 1 gives the year in which the currency used in a particular city had for the …rst time full legal tender status in another state.30 Table 1 gives also the year in which a particular city-market had its earliest bilateral rail connection in our sample. For example, the rail track between the Saxony cities of Dresden and Leipzig was completed in the year 1839, and since this was the earliest connection in the sample for both cities, Table 1 lists this year for Dresden and Leipzig. The trains variable T Rijt for the Dresden-Leipzig pair is 0 until the year 1839, and 1 thereafter. This coding is not based on when a particular city became part of the railway network by getting its railway station, which in itself may not be particularly important. Instead, we code the T R variable speci…c to bilateral connections in our sample. Moreover, since it clearly matters for the choice between di¤erent modes of transport how circuitous the route between two markets is, we have set T Rijt only equal to one once a direct and non-circuitous train connection existed. This has been determined by analyzing historical maps that give the precise geographic location of the historical train tracks in Europe. For example, Figure 2 shows the train connections in the year 1850, as well as the four cities Strassbourg, Munich, Hamburg, and Cologne. In the following year, 1851, the North-South connection between Munich and Hamburg was established, and the variable T Rijt switches from 0 to 1 in our analysis. The T R variable also incorporates other relevant elements of Europe’s topography, such as the existence of bridges across rivers. For example, the railway line between Cologne and 30

The two Alsatian cities of Mulhouse and Strassbourg are special cases, since they were part of France until 1871 and part of Germany from 1871 to 1918. Thus, the value of LTijt between Mulhouse and Toulouse, e.g., goes from 1 to 0 after 1871. Moreover, we could in principle take into account the fact that the uni…ed Germany and other countries in our sample went on the gold standard in the 1870s. We have not done so mainly because being on the same commodity standard is not identical to mutually agreed upon legal tender status. We have also considered the e¤ects of …xing exchange rates on the price gaps. Incorporating this into our analysis does not qualitatively change our …ndings.

20

Aachen was an early one in Europe, completed in the year 1841, and as early as 1843 this line connected internationally to the Belgian cities of Brussels and Brugge. Grain from the relatively low-price areas of Prussia could be shipped via Hanover to the emerging industrial areas of Cologne by the year 1847 via the Köln-Mindener line. But that was only the CologneDeutz part of Cologne, located on the east side of the Rhine— the railway bridge across the Rhine was completed only in the year 1859, and until then, Aachen as well as the Belgian markets could e¤ectively not be supplied by rail with the relatively cheap Eastern European grain.31 Below we also employ information on when states formally abolished serfdom as an indicator of institutional quality. Late abolition of serfdom is taken as a sign of institutions that are not conducive to economic e¢ ciency and growth, consistent with Acemoglu, Cantoni, Johnson, and Robinson (2008) who show that prosperity in 19th century Europe was lower when the prevalence of feudal institutions increased. Such institutions might also a¤ect the e¢ ciency of steam trains. Table A6 of the appendix reports the year of the initial decree abolishing serfdom is presented for our markets. The average year is 1809, with the earliest year being 1783 (Grand Duchy of Baden) and the latest year being 1848 (Austria-Hungary). Finally, we analyze whether state and private railways had di¤erent impacts on trade and the size of the market. On the one hand, private railways might emerge whenever e¢ ciency gains outweigh the cost of adoption, whereas state railway might serve also purposes other than economic e¢ ciency. On the other, there might be market failures, for example imperfect capital 31

We have also experimented with another railway variable that incorporates information on how much freight tra¢ c was present on a given rail line in a given year, based on information in Fremdling et al. (1995). This takes into account di¤erences in the relative importance of rail connections for freight tra¢ c. Moreover, the actual freight tra¢ c …gures also re‡ect di¤erences across rail lines and over time in terms of freight charges per ton-kilometer. Results with this alternative variable were found to be similar.

21

markets, that the state railways might be able to overcome. We have evaluated the historical record to determine whether state or private railways were most important for each of the 68 markets in our sample, and the result is shown also in Table A6 of the appendix.32

Data sources Major sources for the wheat price data are Shiue and Keller (2007) as well as Seu¤ert (1857). The information on trains comes mainly from IEG (2008) and Putzger (1997). The customs liberalization and currency agreement variables are based on the accounts in Henderson (1959), Hahn (1984), and Willis (1896), as well as historical maps at IEG (2008). For population data, we draw on Bairoch et al. (1988), de Vries (1984), Kunz (2008) and Mitchell (1980). The institutional variables are based on Blum (1978) and Fremdling, Federspiel, and Kunz (1995) for the abolition of serfdom and state versus private railways, respectively. Additional details can be found in the appendix. We now turn to the empirical analysis.

4

Empirical Results

To assess the importance of customs liberalization in bringing down price gaps between markets, consider the following regression:

pdifijt =

0

+

1 CUijt

32

0

+ X + "ijt

(1)

The key criteria is whether the state had an important role in …nancing and in operating the railway early on. Since private operators typically needed a concession (license) awarded by the government, the state had always some role in the decision to adopt steam trains. While typically states adopted either private or state railway systems, this is not the case for Prussia and Bavaria, which had mixed systems. Our analysis takes this into account.

22

which relates the log absolute bilateral price gap between markets i and j in year t to the dichotomous customs liberalization variable CUijt and a vector of control variables X. We are interested in estimating

1.

The concern is that customs liberalizations were not exogenous,

so that CUijt is correlated with "ijt and OLS estimates are inconsistent. To address this, we adopt an instrumental variable (IV) approach:

pdifijt =

0

+

1 CUijt

0

+ X + "ijt (2)

CUijt

=

0

+

1 DistCoastij

+

2 ZollP opijt

+ X 0 + vijt

where DistCoastij and ZollP opijt are two instruments for customs liberalization. The …rst captures the distance of markets to the coast. Almost all of the customs liberalizations between cities in this sample were due to the enlargement of the Zollverein. Moreover, the date of a state’s accession to the Zollverein is clearly related to the distance to the coast, with more distant markets joining earlier. Not being a member of the Zollverein mattered more for the states in the South of Germany, since the external tari¤ of the Zollverein prevented customsfree access to the coast, which gave relatively low-transport access to distant markets. It is thus not surprising that by the year 1836, all German states to the south of Prussia had joined the Zollverein, see Figure 3. As one would expect, based on this there is a strong cross-sectional relationship between the distance to the coast and the year of Zollverein accession (R2 of 0.48, see Figure A1 in the Appendix). The bilateral variable DistCoastij equals the minimum of the distance to the coast for market i and for market j. If at least one of the markets is located near the coast, customs between i and j would tend to be not liberalized. To gain precision, we add a second instrument 23

based on the distance-weighted Zollverein population in that particular year, ZollP opijt ; it is de…ned as the log average of the distance-weighted Zollverein populations in market i and j, ZollP opit and ZollP opjt , respectively, where

ZollP opit =

X I ZV st

s S

P opst dis

; 8i; t:

(3)

Here, P opst is the population of state or country s in year t; dis is the geographic distance ZV between market i to the capital of state s, and Ist is an indicator variable that is equal to one

if state s in year t was part of the Zollverein, and zero otherwise. A larger Zollverein population means a larger customs-free internal market in the customs union, and thus the greater is the incentive to join. Note that this variable changes over time while the distance-to-coast variable does not. We adopt analogous IV approaches for trains and currency agreements, with the following instrumental variables. The size of the markets that the railway would connect was an important consideration for rail construction. There is a strong cross-sectional relationship between city population in the year 1800, which pre-dates railway construction anywhere, and the earliest date at which a city-market had a railway connection (shown in the Appendix, Figure A2). This indicates that on average, larger cities adopted railways earlier than smaller cities. The …rst bilateral instrumental variable for T R is the average of the population sizes in city i and city j in the year 1800 (denoted Size1800 ij ). Also the market potentials of locations i and j are employed to predict whether there exists a train connection between them, where the market potential of location i is de…ned as M Pit =

P

s S

P opst , dis

8i; t.

A stylized fact in regional economics is that a location’s market

24

potential is a strong predictor of its economic potential (Harris 1954). Since a city’s market potential is computed from the distance-weighted sizes of all states in the sample, the in‡uence of the city’s size itself on its market potential is negligible. This is relevant for the exclusion restriction, because it reduces the likelihood that the instrument varies systematically with characteristics that determine the price gap between markets i and j: The market potential of markets i and j at time t, M arket_P otentialijt ; is de…ned as the log average of M Pit and M Pjt , for all pairs ij and years t. We construct also two instruments for the currency agreement variable. The …rst is based on the extent to which monetary systems imposed di¤erential transactions costs for trade between market pairs. In the Southern German states, transactions costs were particularly high because coins from di¤erent states had highly varying silver contents even though they had the same face value, one Gulden. These states had relatively more to gain from currency agreements, and the Southern German states formed currency agreements before other states did. The instrument for the bilateral LTijt variable is a dichotomous variable that is equal to one if both markets i and j belong to the Gulden area, and zero if at least one of the markets is not part of the Gulden area (the instrument is denoted by Gulden1754 ). This Gulden area was formed about half a century before the sample period begins, in the year 1754. A second instrument for currency agreements is created as the distance-weighted log average population covered by full legal tender status, denoted by CurrP opijt : It is de…ned as the Zollverein population variable CA above, except that an indicator variable for currency agreement membership, Iijt , plays the ZV role of the Zollverein indicator Iijt in equation (3).33 33

Both of our currency agreement instrumental variables are related to the factors that raise the gains from a currency union stressed by Alesina and Barro (2002). First, the reason why the gains from currency agreements for the Southern German states (of the Gulden1754 bloc) were relatively high had to do with these states …nancing the wars following the French Revolution in ways that were more detrimental to the value of their

25

We employ only a subset of the data in the empirical analysis, namely observations at …veyear intervals (1800, 1805, ..., 1895). Since in the case of wheat, shocks to prices often a¤ect the crop for several years, using annual information would yield relatively little additional information while at the same time creating serial correlation.34 Table 2 presents summary statistics of the data. The average price gap in our sample is 0.15, with a standard deviation of 0.12. The table also shows that the fraction of observations where customs was liberalized is somewhat higher than the fraction where a train connection existed (30% versus 21%, respectively). This re‡ects the fact that the customs liberalizations started in the late 1820s, which is at least a decade before the building of railway tracks gained momentum in Germany. About 13% of the sample had the Gulden currency in the year 1754, and the smaller distance to the coast of the two locations in a market-pair is on average 181 kilometers. The …nal row of Table 2 shows that the average number of observations per market-pair is 11.4, or an equivalent period of 57 years during the 19th century. We now turn to estimating the e¤ect of customs liberalizations, train connections, and currency agreements on bilateral price gaps. currencies than the Middle and Northern German states (see Rittmann 1975, 467-469; Holtfrerich 1993, 521). Thus, the commitment value of currency agreements for the Southern states was relatively high. And the CurrP op variable re‡ects in part the idea that the propensity to join an existing currency agreement rises, the larger is the size of the economy governed by the currency agreement. 34 For the most part, we also restrict our attention to observations that lie in a window of twenty-…ve years before or after the institutional and technological changes that we consider; this gives 6,990 observations. Results from employing all available observations are presented in the Appendix, Table A1.

26

4.1

Market Expansion and the Relative Costs versus Bene…ts of Change

In Table 3, speci…cation (1) presents results for the following two-stage least squares (TSLS) regression pdifijt =

s(i)s0 (j)

+

t

+

1 CUijt

+ "ijt ;

CUijt where

s(i)s0 (j)

=

s(i)s0 (j)

+

t

+

1 DistCoastij

are state-pair …xed e¤ects, and

t

+

2 ZollP opijt

+ vijt

are time …xed e¤ects.35 Our sample includes

68 markets in 20 states and countries. Each of the more than two thousand market pairs ij belongs to a particular state-pair, with the corresponding bilateral …xed e¤ect denoted by 36 s(i)s0 (j) .

These …xed e¤ects control for unobserved heterogeneity at the state-pair level: for

example the geographic distance between states which typically will a¤ect transport costs, and the location of rivers, which may a¤ect the importance of steam trains in bringing price gaps down.37 The …rst three speci…cations in Table 3 are included to assess the performance of the instrumental variables; the …rst-stage results are shown in Panel B, while the second-stage results are presented in Panel A. In column (1), the positive coe¢ cient on the distance to the coast 35

Reported are robust 2-step e¢ cient GMM standard errors, which are clustered at the state-pair level. This is preferred since customs liberalizations typically occur at the state-pair level, so that trade barriers between any markets for a given state-pair often fall away in the same year. The analogous is often the case for currency agreements as well. Also train connections were frequently established so that the opening of a particular train connection also connected other markets for the same pair of states. 36 With 20 states and countries, there are up to 400 state-pair …xed e¤ects. In the regressions, we estimate close to 300 of those (not reported); the remaining ones cannot be estimated since such state-pairs are not observed in the sample. 37 For some of the larger states, such as Prussia, state pair …xed e¤ects imperfectly control for bilateral distance between markets. We address this by showing that the results are similar using market-pair instead of state-pair …xed e¤ects. The regressions are also weighted by the number of bilateral price gap observations for a particular pair; unweighted regressions yield quite similar results. See Table A2 in the appendix.

27

indicates that markets in states that are relatively far away from the seaboard joined the Zollverein relatively early. This was expected, given the strong positive cross-sectional relationship mentioned earlier, but the …rst-stage regression produces this result controlling for unobserved heterogeneity. We also …nd that Zollverein membership became more attractive as the Zollverein population grew (coe¢ cient of 0:022 on ZollP op). In the train regression of column (2), the market potential and the population in the year 1800 enter positively. This is consistent with the idea that both raise the likelihood that a train connection between markets exists, though only the market potential variable is signi…cant at standard levels. In the currency agreement speci…cation (3), being in the Gulden area early on (Gulden1754 ) is a good predictor of whether currency agreements exist, and moreover, the larger is the population covered by currency agreements, the more likely it is that additional states join the agreement. To sum up, the instrumental variables have the expected impact. What about the power of the instruments? Given the poor performance of instrumental variable estimation with weak instruments (Bound, Jaeger, and Baker 1995, Staiger and Stock 1997), we use two statistics to gauge the power of the instruments. Table 3 reports the p-values of the F-test of the excluded instruments, as well as Shea’s partial R2 for the excluded instruments. The R2 values are relatively low, ranging from about one to three percent. Moreover, the p-value for the F-test of the train …rst-stage is one percent, which also suggests that power is limited.38 However, it is plausible that there is variation in these IV e¤ects. France, for example, was never o¤ered Zollverein membership, and thus it is not clear that the Zollverein population has the same impact on the propensity of France to join as it does for German states. Thus, we allow the instrumental variables’in‡uence to vary by state. The equations for the customs e¤ect then 38

This F-statistic is 5.25, below the rule of thumb of a minimum of 10 suggested by Staiger and Stock (1997).

28

become pdifijt =

s(i)s0 (j)

+

t

+

1 CUijt

+ "ijt ;

CUijt where

1s

and

2s

=

s(i)s0 (j)

+

t

+

1s DistCoastij

+

2s Zoll_P opijt

+

(4)

ijt

are the …rst-stage parameters varying by state.39 The results are shown

in columns (4), (5), and (6) of Table 3.40 The additional instruments raise the explanatory power of the …rst stage by at least a factor of three, and also the F-tests’p-values are now all lower than 0.001. The possibility that our inferences are strongly a¤ected by weak instruments appears to be remote now. Looking at the second-stage results, one sees that customs liberalization, trains, and currency agreements all increase market size as evidenced by lower price gaps. The trains e¤ect is the largest, at about -0.22, followed by the customs impact at about -0.13, but also the currency estimate is substantial, at about -0.10. These results are preliminary since the relative impact of institutional and technological change can best be gauged by including both into the same regression, something we will turn to below. For now, it is instructive to compare the IV results with the corresponding OLS results, which are shown in Panel C of Table 3. According to the OLS estimates, neither customs liberalization nor currency e¤ects have a signi…cant impact on price gaps, which is in sharp contrast to the IV estimates. Moreover, even though both the OLS and the IV estimate point to a signi…cant price-gap reducing e¤ect from steam trains, the IV estimate is more than eight times as large as the OLS estimate. It is thus not surprising that for all three variables speci…cation tests strongly reject the null 39

With S = 20 states and countries, the maximum number of instruments here is 40. The actual number of instruments that are employed depends on the available price data; it is listed in the table. 40 From now on, the …rst-stage estimates ^1s and ^2s are not shown to conserve space; they are available from the authors upon request.

29

hypotheses of exogeneity (second to last line in Table 3).41 This is the …rst main result of our paper: accounting for the endogeneity of institutional and technological change is crucial for the estimation results. There are a number of key questions. First, why are our OLS estimates relatively close to zero whereas others have recently found sizable impacts in OLS regressions? One reason might be that in our sample, there are many factors that a¤ect institutional and technological change, and they outweigh each other to a substantial degree.42 Second, why are our IV results larger (in absolute value) than the OLS estimates? Take the trains impact, for example. If endogeneity of train adoption induces selection such that those market-pairs expecting large price gap reductions adopt trains before other market-pairs anticipating smaller reductions, the IV estimate would be smaller (in absolute value), not larger than the OLS estimate. However, this is not what we …nd. Instead, the results suggest that many of the late train-adopting market-pairs are those that would have bene…ted relatively strongly, and those late-adopters tend to be relatively small cities. In that case, the relatively large e¤ect of trains on shrinking price gaps becomes apparent only once the IV estimation un-does the relation between timing of adoption and market size. In addition, the IV estimation might also address measurement error issues that could bias the OLS estimate towards zero. Speci…cally, the 0/1 trains variable 41

These Hausman-style tests are based on the di¤erence between IV and OLS estimates. Since only IV is consistent in the presence of endogeneity, a large di¤erence between IV and OLS results supports endogeneity. The large di¤erence between OLS and IV results also gives another perspective on the possibility of weak instruments. Weak instruments would bias the IV results towards the OLS estimates (Bound, Jaeger, and Baker 1995). Given the substantial di¤erences between OLS and IV estimates in (4) to (6), that appears to be unlikely. 42 We know from the historical record, for example, that personal animosity between rulers a¤ected the likelihood of institutional change in form of currency or trade agreements between them. Also, the leader of some states were more interested in steam trains than others. See Henderson (1959), Hahn (1984), and Fremdling, Federspiel, and Kunz (1995) on this. Another reason for the OLS estimates close to zero might lie in the bilateral …xed e¤ects we include, which goes beyond sets of importer and exporter …xed e¤ects that are often employed.

30

is likely measuring the trains impact with error, since as noted above the trains e¤ect would have been dependent also on the train freight tari¤ structure, among other factors. Similarly, the actual Zollverein accession date of certain states was occasionally a¤ected by the personal preferences of the states’leaders.43 Because the IV estimation, in contrast, predicts Zollverein accession with the distance to the coast and Zollverein population, it is not prone to this type of measurement error. Another important issue is whether the instruments are exogenous to price gaps. As usual, there can be no direct test of the exclusion restrictions. Exogeneity of the instruments is however plausible for a number of reasons. First, on a priori grounds, DistCoast is based on exogenous geographic factors while Gulden1754 and P op1800 are based on pre-sample information. And as noted above, the market potential, Zollverein population, and currency agreement population measures by construction do not depend critically on characteristics of the market-pair ij itself. Second, we have employed overidenti…cation tests that ask whether the instruments as a set appear to be valid. The p-value of Hansen’s J test statistic is shown in the last row of Table 3, and in none of these tests can we reject the null of instrument validity at standard levels of signi…cance. Third, we have conducted some informal analysis on the exclusion restrictions by computing the sample correlation between price gap, the dependent variable, and the instruments. These correlations are typically quite low. For example, the correlation of pdif with P op1800; DistCoast; and Gulden1754 is equal to

0:017; 0:013; and

0:024; respectively. Moreover, in exploratory regressions of the price gap on the instrumental variables, the latter are typically not signi…cant. 43

Or, for that matter, by the preferences of foreign leaders. The decision of Hanover to delay its accession to the Prussian-led Zollverein, for example, was to a signi…cant extent determined by the preferences of the King of England, who ruled Hanover in personal union.

31

The following section turns to estimating the impact of institutional and technological change in direct comparison to each other.

4.2

Institutional and technological change compared

We …rst consider customs liberalizations and the establishment of train connections. In Table 4, column (1), the customs e¤ect is estimated at about 7% and the trains e¤ect at about 13%. These estimates are lower by about 40% compared to the results in Table 3, where one variable at a time is included. This suggests that, quite plausibly, customs liberalization and train adoption explain to some extent the same variation in changing price gaps. The …rst-stage regressions are strong and the Hansen J test indicates that exogeneity of the instruments cannot be rejected at standard signi…cance levels. Comparing the impact from currency agreements with the introduction of steam trains, the former lowers price gaps by about 6% while trains have a 16% e¤ect (column (2)). How about the impact of trains relative to both customs liberalization and currency agreements? It turns out that it is di¢ cult to estimate jointly the e¤ect from customs liberalization and currency agreements, since markets that bene…ted from one typically also bene…ted from the other (the correlation of CU and LT is 0.75). In order to be able to compare the impact from steam trains with a broad institutions e¤ect, we construct a new variable, IN STijt that incorporates both currency and customs information:

IN STijt =

8 > > > 0 if CUijt = 0 and LTijt = 0 > > >
> > > > > : 2 if CUijt = 1 and LTijt = 1 32

(5)

Speci…cation (3) in Table 4 gives an impact for this institutions variable of about 3%, while the trains e¤ect is estimated at about 13%. The estimate of 3% for IN ST might appear somewhat low in the light of 7% for customs and 6% for currency agreements. At the same time, given that CU and LT are correlated, one would expect that CU in column (1) and LT in column (2) pick up part of the e¤ect from the other institutional variable, and, if one compares the CU and LT estimates for the impact for IN ST = 2 (both CU and LT ), they are similar, namely 6%. Moreover, also diminishing returns to institutional change might explain part of the …nding. To sum up, with train estimates of around 14% and customs or currency impacts of about 6.5%, the impact of technological change on market size in the 19th century was about twice as large as that of institutional change. This is our second major …nding. We now turn to a number of important robustness checks, see Table 5. While the baseline results from Table 4 are repeated on the left in columns (3) and (4), the sample is restricted to the years from 1820 to 1880. It is during these 60 years that most of the train connections were established, and when customs borders were liberalized, and currency agreements were formed. Moreover, this period also excludes the early 1800s, where prices and trade may be particularly strongly a¤ected by wars (the Napoleonic Wars ended in 1815). Next, we show results from trimmed price gap samples. Speci…cally, there we drop the observations that exhibit the 2.5% highest and 2.5% lowest price di¤erences during the sample period. This enables us to see whether our results are strongly driven by a small number of unusual but in‡uential observations. We see that for the years 1820-80, the impact of institutions is slightly larger than in the baseline, and for the trimmed price gap sample, it is somewhat smaller, but overall these results are not very di¤erent. We also examine the robustness of the results in terms of focusing on the customs lib-

33

eralizations associated with Zollverein accession treaties, because there were other customs liberalizations, and these might a¤ect our results through direct or third-country e¤ects. First, there were customs liberalizations among German states outside of the Zollverein liberalizations. Second, the Zollverein as a whole liberalized trade for some years with non-Zollverein countries. And third, countries outside of the Zollverein at times liberalized trade between each other.44 While our analysis incorporates these to some extent, we are far from having complete information on grain protection and liberalization for this 19th century sample.45 However, the single biggest event in this respect occurred in the third quarter of the 19th century, when many countries liberalized their trade. We know that the Zollverein had no external duties on wheat for some time after the year 1853 (Tracy 1989, 87; Henderson 1959, 226). Only with the arrival of grain from the United States about two decades later, pressure for import protection mounted and in 1879, the German Reichstag reverted to import tari¤s for wheat (Tracy 1989, 89). In Table 5, we compare the results between our baseline and two alternative treatments of customs liberalizations. For the "Pervasive Liberalization" speci…cation (7), we assume that the other European countries’policies were identical to that of the Zollverein, respectively the German Reich; we assume that during the years 1853 to 1879, no customs duties existed between any two markets in the sample. For the "Only ZV Liberalization" speci…cation (8), we make the alternative assumption that there were no external customs duties for the Zollverein markets, 44

An example of the …rst is the South German Customs Union between Bavaria and Wurttemberg. It was formed in the year 1828 and lasted until 1833, when both states became part of the Zollverein. Our analysis incorporates the direct but no third-country e¤ects. An example of Zollverein customs liberalizations with other countries is the agreement with Belgium in the year 1838. However, these were neither as comprehensive nor as long-lasting as the Zollverein liberalizations. An example of the third point are the customs liberalizations throughout Europe, as discussed in the following. 45 For general information on 19th century trade agreements, see Pahre (2008).

34

and that tari¤s greater than zero remained in place between the non-Zollverein markets.46 The results for speci…cations (7) and (8) indicate that either treatment leads to similar conclusions as our baseline. This suggests that accounting fully for temporary liberalizations and controlling for third-country e¤ects will not change our main …ndings. We have also extended this robustness analysis in other dimensions.47 This can be found in Tables A1 to A4 in the Appendix. Overall, these results con…rm the above …ndings.

4.3

The direct and indirect e¤ect of institutions

In this section we expand the analysis to consider indirect e¤ects of institutions through their impact on the introduction of steam trains. Figure 4 illustrates the idea: in addition to the direct e¤ect of institutions on market size–the solid arrows from currency agreements and customs liberalizations–we now extend our analysis to indirect e¤ects from institutions on the adoption of steam trains on market size (the dashed arrows). We consider the interaction of two institutions on the adoption of steam trains, one more proximate and one more fundamental. The former is whether the steam trains for a particular market were primarily run by the state or by private agents. There is a general tendency in the literature to assume that private railways would be more e¢ cient (e.g., Fremdling 1975), although as noted above this need not be so if the state provides an important public good. In 19th century Central Europe, both state railways and private railways were signi…cant in size, which enables us to compare the impact of the two on market size. 46

Even though many European countries had relatively low duties on grain in the period of 1850-80, and in particular between 1866-79, they were typically not equal to zero; see Bairoch (1989) and Jacks (2005). 47 This includes shorter and longer time horizons (instead of a 25 year window), limited-information maximum likelihood estimation (instead of TSLS), additional controls for unobserved heterogeneity (market-pair …xed e¤ects instead of state-pair …xed e¤ects), unweighted regressions, and robustness analysis by country.

35

We also use a more fundamental institutional variable, which is the date at which serfdom was formally abolished in a state or country. We take the abolition of serfdom as a general sign that highly discretionary and exploitative activities by the state towards private agents are becoming less likely.48 To the extent that this general institutional setting a¤ects the impact of steam trains on market size, we expect that states that abolish serfdom relatively early will tend to see stronger gains from train adoption than late-abolishing states. It is important to note that in this analysis we treat the existence of a particular form of railway ownership and serfdom as exogenously given.49 The results are shown in Table 6. In speci…cations (1) and (2), we add two variables to the baseline results: …rst, there is T rain_StateRailway, which is the interaction of the 0/1 trains variable with StateRailways, an indicator of the importance of state railways in market pair ij. The latter is a constant for each market pair ij; taking on values between 0 (both markets are private) and 1 (both markets have state railways). This variable is also included by itself to avoid misspeci…cation. In the regression, we treat the interaction variable T rain_StateRailway as endogenous and StateRailways as exogenous. The results suggest that state-run steam trains had a signi…cantly smaller impact on market size than private-run railways. Moreover, with around 70 percent, the e¢ ciency of state railways in reducing price gaps is substantially lower.50 48

It is not crucial for our analysis to know when serfdom in a particular state or country fell out of use (which was typically considerably earlier than the time when it was formally abolished). 49 We return to this issue in section 5. Note that from Figure 4, one might also think that the existence or non-existence of serfdom might in‡uence the impact of currency agreements and customs liberalization in an economy. We have explored this possibility but found little evidence for it. Further, one might postulate that serfdom (or state railway ownership) would explain the timing of when train connections were built. However, since the correlation of the 0/1 trains variable with either the date of the abolition of serfdom or whether state railways dominated is low, this is unlikely. 50 A change from StateRailway = 0 to StateRailway = 1 is associated with a 85% lower trains e¤ect in speci…cation (1), and with a 55% lower impact according to speci…cation (2).

36

In columns (3) and (4), we include another interaction variable T rain_LateAbolition. Here, LateAbolition is a 0/1 variable taking the value of 1 if serfdom was present in at least one of the markets i and j after the year 1831. The rationale of using the maximum year in pair ij is that low institutional quality in one market is enough for it to be unlikely that trains will be important in leading to trade. We …nd that the impact of trains on market size tends to be lower in states that abolished serfdom relatively late, compared to states that abolished serfdom earlier. The …ndings are consistent with theories in which the establishment of non-absolutist institutions (no serfdom), or reliance on private railways, lead to greater increases in market size than when railways are state-operated, or the state is feudal. These results are broadly con…rmed in a number of robustness checks (see Table A5 in the appendix). At the same time, one needs to be cautious in interpreting these results. First, the size of the reduction in the trains e¤ect from late-serfdom-abolishers is smaller than that from state railways (20 versus 70 percent). Second, the coe¢ cient on the linear LateAbolition variable in (3), as well as on StateRailway in (1) and (2), is negative. Relatively low price gaps among late-abolishing market-pairs (or with state-run railways) is not what one expects if early abolishment of serfdom is a general sign of economic e¢ ciency-oriented institutions. In addition, the simple correlation between StateRailway and LateAbolition is negative, that is, state railways dominated where serfdom was abolished relatively early. It appears, therefore, that while there is evidence that the technology e¤ect of railways indeed is in‡uenced by the speci…c institutional setting, more work is needed to determine the precise relationship between proximate and fundamental indicators of institutional quality on the one, and their relationship with technological change on the other hand. The following section presents some concluding discussion.

37

5

Conclusions

Do institutional and technological change a¤ect the size of the market? To answer this question, we examined systematic deviations from the Law of One Price in Western and Central Europe over the 19th century. This setting is particularly interesting since at that time national and international markets were …rst emerging. To examine systematic deviations from the Law of One Price, we focus on the market price of wheat in cities in large parts of Europe. The analysis embodies three determinants of trade and the size of the market: customs liberalizations, currency agreements, and train transportation. Admittedly, this list may be incomplete. However, the analysis covers the Zollverein liberalizations, monetary uni…cation in Germany, and transportation improvements in form of steam trains, which are a priori all of …rst-order importance for European economic development during this time. Our empirical results strikingly demonstrate that both institutional and technology change must …gure prominently for any understanding of the expansion of markets and economic development. It would have been impossible to arrive at these results without extensive data. To this end, we constructed a unique database on wheat prices, geographic, city, state, and country characteristics, customs and currency agreements, and train connections for 68 markets in a major part of Europe over the entire 19th century. This detailed data lets us address the reality that institutional as well as technological change are endogenous responses to economic conditions at the time. We …nd that accounting for this endogeneity gives vastly di¤erent results than not doing so. Quantitatively, while there is some variation our estimates are for the most part highly consistent. We …nd that the introduction of steam trains reduced price gaps by about fourteen

38

percentage points; customs liberalizations lowered price gaps by about seven percentage points and currency agreements by about six percentage points. Since there appears to be diminishing returns, the combined impact of currency agreements and customs liberalization for the typical market would still be substantially below what the introduction of steam trains would do. Thus, we …nd that technological change had a larger e¤ect on market size than institutional change in 19th century Europe. With a decline in price gaps of about twenty-two percentage points in the sample, the introduction of trains together with currency and customs agreements account for most of the overall decline in price gaps over this period. We have also asked whether there exist additional indirect e¤ects from a state’s institutions. In particular, we were interested to see whether a state’s institutions in‡uence the rate of technological change, in this case the adoption of steam trains. There are a number of possible channels of why this might be so, and correspondingly, the paper examines whether either proximate or more fundamental institutions have played a role. Both the former, measured by the ownership of railways, as well as the latter, captured by whether serfdom existed, are found to have in‡uenced the e¢ ciency of steam trains in raising the size of the market. In particular, for markets that were served primarily by state railways, or that were located in states that abolished serfdom relatively late, the introduction of steam trains reduced price gaps by subtantially less than when railways were privately run or when serfdom had been abolished early. Thus our analysis identi…es an important indirect e¤ect from institutions on economic performance. This research suggests a number interesting areas of future research. One is the relationship between institutions and technological change. Recently, economists have been quite successful in explaining which countries are rich and which are poor with variation in institutional quality.

39

There is also plenty of evidence that institutional quality a¤ects the rate of technical change.51 At the same time, we still know too little about what exactly determines whether an economy’s institutions will foster technological change, and how fundamental institutions are linked to more proximate ones in their e¤ect on economic e¢ ciency and growth. The analysis in this paper is only a …rst step in this direction. Our results also indicate that studies of the impact of technological change today might have a very high return. One example is the impact of advances in information and communications technologies— the transactions costs of technological knowledge— on o¤shoring and the associated relocation of both jobs and production, as well as the resulting changes in global trade ‡ows. The analysis in this paper suggests more generally that with a good measure of technological change, it is not only feasible to estimate the impact of technology on economic outcomes, but such studies may also yield dramatic new insights on the economics of development.

51

Recent work includes Acharya and Keller (2008) and Coe, Helpman, and Ho¤maister (2008).

40

References [1] Acemoglu, D., D. Cantoni, S. Johnson, and J. Robinson (2008), "From Ancien Regime to Capitalism: The Spread of the French Revolution as a Natural Experiment", Harvard and MIT, July 2008. [2] Acemoglu, D., S. Johnson, and J. Robinson (2005), "Institutions as the Fundamental Cause of Long-run Growth", Handbook of Economic Growth, P. Aghion and S. Durlauf (eds.), Elsevier 2005. [3] Acemoglu, Daron, Simon Johnson and James A. Robinson (2001) “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review, December, 91, 5, 1369-1401. [4] Acharya, R., and W. Keller (2008), "Estimating the productivity selection and technology spillover e¤ects from imports", NBER Working Paper # 14079. [5] Alesina, Alberto and Robert J. Barro (2002). “Currency Unions”, Quarterly Journal of Economics, May, 409-436. [6] Baier, Scott and and Je¤rey Bergstrand (2007), “Do Free Trade Agreements Actually Increase Members’International Trade?”, Journal of International Economics. [7] Bairoch, Paul (1989), "European trade policy, 1815-1914", Chapter 1 in The Cambridge Economic History of Europe, Vol. 8, Peter Mathias and Sidney Pollard (eds.), Cambridge University Press: Cambridge, pp. 1-160. [8] Bairoch, Paul, Jean Batou, and Pierre Chevre (1988), The Population of European Cities from 800 to 1850: Data Bank and Short Summary of Results. Geneva: Librairie Droz. 41

[9] Barro, R., and S. Tenreyo (2004), "Economic E¤ects of Currency Unions", Economic Inquiry 45, 1: 1-23. [10] Blum, J. (1978), The End of the Old Order in Rural Europe, Princeton University Press: Princeton, NJ. [11] Bound, J., D. A. Jaeger, and R. Baker (1995), "Problems with Instrumental Variables Estimation when the Correlation between the Instruments and the Endogenous Explanatory Variable is Weak", Journal of American Statistical Association, Volume 90, No. 430, 443-450. [12] Braudel, Fernand (1992), The Wheels of Commerce, Volume 2. Berkeley: University of California Press. [13] Cameron, R., and L. Neal (2003), A Concise Economic History of the World, 4th edition, Oxford University Press: New York and Oxford. [14] Coe, D., E. Helpman, and A. Ho¤maister (2008), "International R&D Spillovers and Institutions", NBER Working Paper # 14069. [15] De Vries, Jan (1984), European Urbanization 1500-1800, Cambridge, Massachusetts: Harvard University Press. [16] Dumke, Rolf H. (1976), “The Political Economy of German Economic Uni…cation: Tari¤s, Trade and Politics of the Zollverein Era”, Ph.D. Dissertation, University of WisconsinMadison. [17] Eichengreen, Barry and Douglas Irwin (1995), “Trade Blocs, Currency Blocs, and the Reorientation of World Trade in the 1930s”, Journal of International Economics: 1-24. 42

[18] Estevadeordal, Antoni, Brian Frantz, and Alan Taylor (2003), “The Rise and Fall of World Trade, 1870-1939”, Quarterly Journal of Economics, 118(2): 359-407. [19] Fishlow, Albert (1965), American Railroads and the Transformation of the Ante-Bellum Economy. Cambridge, MA: Harvard University Press. [20] Flandreau, Marc and Mathilde Maurel (2001), “Monetary Union, Trade Integration, and Business Cycles in 19th Century Europe: Just Do It”, Centre for Economic Policy Research, Discussion Paper Series No. 3087, November. [21] Fogel, R.W. (1964), Railroads and American Economic Growth: Essays in Econometric History. Johns Hopkins Press: Baltimore, Maryland. [22] Frankel, Je¤rey, and Andrew Rose (2002), “An Estimate of the E¤ect of Common Currencies on Trade and Income”, Quarterly Journal of Economics. [23] Fremdling, R. (1975), Eisenbahnen und deutsches Wirtschaftswachstum 1840-1879, Dortmund. [24] Fremdling, R., R. Federspiel, and A. Kunz (1995), (eds.), Statistik der Eisenbahnen in Deutschland 1835-1989, Scripta Mercaturae Verlag: St. Katharinen. [25] Fremdling, R., and G. Hohorst (1979), “Marktintegration der preussischen Wirtschaft im 19. Jahrhundert— Skizze eines Forschungsansatzes zur Fluktuation der Roggenpreise zwischen 1821 und 1865”, in R. Tilly and R. Fremdling (eds.), Industrialisierung und Raum, Stuttgart, 56-101. [26] Gerhard, H.-J., and K. H. Kaufhold (1990), (eds.), Preise im vor- und frühindustriellen Deutschland, Göttingen. 43

[27] Glick, R., and A. Taylor (2008), "Collateral Damage: Trade Disruptions and the Economic Impact of War", forthcoming, Review of Economics and Statistics. [28] Goethe (1828), Johann Wolfgang von Goethe: Conversation with Eckermann on Oct. 23, 1828; on the internet at http://www.zeno.org (accessed February 19, 2008). [29] Hall, R. E., and C. Jones (199), "Why Do Some Countries Produce So Much More Output Per Worker Than Others?", Quarterly Journal of Economics 114, 1: 83-116. [30] Hahn, H.-W. (1984), Geschichte des Deutschen Zollvereins, Vandenhoeck & Ruprecht: Göttingen. [31] Hanauer, A. (1878), Etudes Economiques sur L’Alsace Ancienne et Moderne, Vol. 2, Paris. [32] Harley, C. Knick (1988), "Ocean Freight Rates and Productivity, 1740-1913: The Primacy of Mechanical Invention Rea¢ rmed", Journal of Economic History 48: 851-876. [33] Harris, C. (1954), "The Market as a Factor in the Localization of Industry in the United States", Annals of the Association of American Geographers 64: 315-348. [34] Hausman, J. (1978), "Speci…cation Tests in Econometrics", Econometrica 46, 6: 1251-1271. [35] Helpman, E. (2008), "Introduction", in Institutions and Economic Performance, E. Helpman (ed.), Harvard University Press, Cambridge, MA. [36] Henderson, W. O. (1959), The Zollverein, Cambridge University Press, Cambridge, UK. [37] Holtfrerich, Carl-Ludwig (1993), “Did monetary uni…cation precede or follow political uni…cation of Germany in the 19th century”, European Economic Review 37: 518-524.

44

[38] Holtfrerich, Carl-Ludwig (1989), “The monetary uni…cation process in 19th century Germany: Relevance and lessons for Europe today”, in M. de Cecco and A. Giovannini, (eds.), A European central bank? Perspectives on monetary uni…cation after ten years of the EMS. Cambridge: Cambridge University Press. [39] Hummels, D. (2007), "Transportation Costs and International Trade in the Second Era of Globalization", Journal of Economic Perspectives 21, 3: 131-154. [40] IEG (2008), Server für digital Karten am Institut für Europäische Geschichte, Mainz; http://www.ieg-maps.uni-mainz.de/; accessed November 2008. [41] Jacks, David (2006), “What drove 19th century commodity market integration?”, Explorations in Economic History 43: 383-412. [42] Jacks, David (2005),"Intra- and International Commodity Market Integration in the Atlantic Economy, 1800-1913", Explorations in Economic History 42: 381-413. [43] Jacks, D., and K. Pendakur (2008), "Global Trade and the Maritime Transport Revolution", NBER Working Paper # 14319. [44] Kopsidis, Michael (2002), “The Creation of a Westphalian Rye Market 1820-1870: Leading and Following Regions, an Co-Integration Analysis", Jahrbuch für Wirtschaftsgeschichte: 85-112. [45] Köttgen, August (1890), Studien über Getreideverkehr und Getreidepreise in Deutschland. Jena: Verlag von Gustav Fischer.

45

[46] Kunz, A. (2008), eKompendium-HGISG: Kompendium zum Historischen Informationssystem der deutschen Staatenwelt, University of Mainz, Germany, http://www.hgisgekompendium.ieg-mainz.de, accessed March 2008. [47] La¤ut, Michel (1983), “Belgium”, in Patrick O’Brien, ed. Railways and the Economic Development of Europe, 1830-1914. St. Martin’s Press: New York. [48] Lopez-Cordova, J. Ernesto and Christopher Meissner (2003), “Exchange-Rate Regimes and International Trade: Evidence from the Classical Gold Standard Era,”American Economic Review, 93 (1). [49] Milward, A.S. and S.B. Saul (1977), The Development of the Economies of Continental Europe, 1850-1914. Harvard University Press: Cambridge, MA. [50] Mitchell, B. R. (1980), European Historical Statistics, 1750-1975, 2nd edition, Facts on File: New York. [51] North, D. (1990), Institutions, Institutional Change, and Economic Performance, Cambridge University Press, New York. [52] O’Brien, Patrick (1983), “Transport Development in Europe, 1789-1914”, in Railways and the Economic Development of Western Europe, 1830-1914, edited by P. O’Brien. [53] O’Rourke, Kevin, and Je¤rey G. Williamson (1999), Globalization and History, Cambridge: MIT Press. [54] O’Rourke, K., and J. G. Williamson (1994), "Late Nineteenth-Century Anglo-American Factor-Price Convergence: Were Heckscher and Ohlin Right?", Journal of Economic History 94,4: 892-916. 46

[55] Pahre, Robert (2008), Trade Agreements Database, University of Illinois. [56] Putzger (1997), "Entwicklung des Eisenbahnnetzes", in Putzger Historischer Weltatlas, (ed.) W. Leisering, 102nd edition, Cornelsen: Berlin. [57] Ritschl, A., and N. Wolf (2003), "Endogeneity of Currency Areas and Trade Blocs: Evidence from the Inter-War Period", CEPR Discussion Paper No. 4112, November. [58] Rittmann, Herbert (1975), Deutsche Geldgeschichte 1484-1914, München: Battenberg. [59] Rodrik, D., A. Subramaniam, and F. Trebbi (2004), "Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development", Journal of Economic Growth 9, 2: 131-165. [60] Rose, Andrew (2004), “Do We Really Know that the WTO Increases Trade?” American Economic Review. [61] Rose, Andrew (2000), “One Money, One Market: Estimating the E¤ect of Common Currencies on Trade”, Economic Policy. [62] Seu¤ert, G. K. L. (1857), Statistik des Getreide- und Viktualien-Handels im Königreiche Bayern mit Berücksichtigung des Auslandes, J. G. Weiss, Munich. [63] Shiue, C. H., and W. Keller (2007), "Markets in China and Europe on the Eve of the Industrial Revolution", American Economic Review, September. [64] Staiger, D., and J. Stock (1997), "Instrumental Variable Estimation with Weak Instruments", Econometrica 65: 557-586.

47

[65] Subramanian, Avind and Shang-Jin Wei (2007), “The WTO Promotes Trade, Strongly, but Unevenly,”Journal of International Economics 72: 151-175. [66] Tracy, Michael (1989), Government and Agriculture in Western Europe 1880–1988, 3rd edition, New York: New York University Press. [67] Vierteljahrshefte (1935), Vierteljahrshefte zur Statistik des Deutschen Reichs, Year 1935, Germany, Statistisches Reichsamt. [68] Viner, Jacob (1950), The Customs Unions Issue, New York: Carnegie Endowment for International Peace. [69] von Eisenhart Rothe, Wilfried, and Anton Ritthaler (1934), Vorgeschichte und Begründung des deutschen Zollvereins : 1815-1834, Berlin 1934. [70] von Viebahn, Georg (1862), Statistik des zollvereinten und nördlichen Deutschlands, Zweiter Theil (Bevölkerung, Bergbau, Bodenkultur), Georg Reimer Verlag, Berlin 1862. [71] Williamson, Je¤rey G. (1980), “Greasing the Wheels of Sputtering Export Engines: Midwestern Grains and American Growth”, Explorations in Economic History, 17: 189-217. [72] Willis, Henry Parker (1896), “The Vienna Monetary Treaty of 1857”, Journal of Political Economy, pp.187-207.

48

Table 1: Summary Statistics Overall sample period: 1800 - 1899

No

City

State/Country

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

Prague Salzburg Venice Vienna Baden Augsburg Bamberg Bayreuth Erding Kempten Landshut Lindau Memmingen Munich Noerdlingen Nurnberg Regensburg Straubing Wuerzburg Zweibruecken Brugge Brussels Braunschweig Bar-le-Duc Chalons sur Marne Luneville Mulhouse Strassburg Toulouse Bremen Frankfurt/Main Hamburg Luebeck

Austria-Hungary Austria-Hungary Austria-Hungary Austria-Hungary Baden Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Bavaria Belgium Belgium Brunswick France France France France France France Free City Free City Free City Free City

Number of price obs.

Mean price

Year of Earliest Obs.

8 4 7 86 28 41 41 41 41 41 41 41 41 100 41 45 41 41 41 38 100 91 50 30 30 30 76 76 100 11 14 100 9

19.47 29.02 15.57 20.57 16.29 16.92 16.32 16.82 16.33 18.81 15.58 19.14 18.00 18.69 16.14 16.42 15.09 14.65 16.41 16.57 20.62 22.45 16.50 18.08 18.55 19.03 22.41 21.63 21.40 20.53 22.57 19.68 17.58

1836 1849 1836 1820 1818 1815 1815 1815 1815 1815 1815 1815 1815 1800 1815 1811 1815 1815 1815 1818 1800 1800 1800 1825 1825 1825 1800 1800 1800 1837 1816 1800 1837

Year of Zollverein Accession

1836 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834

1841

1888 1836 1888 1867

Year of Earliest Rail Connection 1845 1860 1856 1845 1846 1840 1844 1853 1859 1852 1854 1852 1858 1840 1849 1844 1859 1858 1854 1857 1838 1838 1844 1851 1851 1851 1841 1841 1859 1847 1840 1846 1851

Year of Earliest Legal Tender Status

1837 1837 1837 1837 1837 1837 1837 1837 1837 1837 1837 1837 1837 1837 1837 1837

1857

1871 1837 1871 1871

Table 1, cont'd

No

City

State/Country

34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68

Goettingen Hannover Kassel Bingen Giessen Mainz Worms Wiesbaden Grabow Boizenburg Parchim Rostock Schwerin Wismar Nijmegen Utrecht Aachen Berlin Cologne Hamm Herdecke Minden Muenster Saarlouis Soest Wetzlar Xanten Dresden Leipzig Zwickau Basel Lucerne Rorschach Stuttgart Ulm

Hannover Hannover Hesse-Cassel Hesse-Darmstadt Hesse-Darmstadt Hesse-Darmstadt Hesse-Darmstadt Hesse-Nassau Mecklenburg Mecklenburg Mecklenburg Mecklenburg Mecklenburg Mecklenburg Netherlands Netherlands Prussia Prussia Prussia Prussia Prussia Prussia Prussia Prussia Prussia Prussia Prussia Saxony Saxony Saxony Switzerland Switzerland Switzerland Wurttemberg Wurttemberg

Number of price obs. 68 50 27 1 1 3 1 1 71 71 71 71 71 57 93 15 61 61 100 20 20 13 64 20 20 20 20 21 68 21 10 9 14 5 6

Prices in Bavarian Gulden, per Bavarian Scheffel (about 223 liter)

Mean price

Year of Earliest Obs.

Year of Zollverein Accession

Year of Earliest Rail Connection

17.12 17.81 14.22 20.34 19.12 23.68 20.68 18.13 18.45 18.30 17.43 17.57 17.67 16.65 21.46 30.66 18.88 18.14 18.25 20.86 23.23 21.49 18.91 17.70 17.71 19.27 18.48 16.78 20.15 18.44 24.75 23.94 20.79 23.68 22.81

1800 1801 1822 1840 1840 1840 1840 1840 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1800 1832 1832 1832 1845 1845 1824 1850 1850

1854 1854 1831 1828 1828 1828 1828 1836 1867 1867 1867 1867 1867 1867

1854 1844 1849 1858 1850 1853 1853 1840 1846 1846 1880 1850 1847 1848 1856 1856 1841 1841 1841 1847 1848 1847 1848 1858 1850 1862 1880 1839 1839 1845 1844 1856 1856 1850 1850

1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834 1834

1834 1834

Year of Earliest Legal Tender Status 1857 1857 1857 1837 1837 1837 1837 1837 1871 1871 1871 1871 1871 1871

1857 1857 1857 1857 1857 1857 1857 1857 1857 1857 1857 1857 1857 1857

1837 1837

Table 2: Sample Summary Statistics

Variable

Mean

Std. Dev.

Variable Description

Price Gap

0.15

0.12

Absolute value of the log difference of wheat price in market i and market j (pdif)

Train Connection

0.21

0.41

0/1 variable; 1 if train connection exists between markets i and j in year t, 0 otherwise (TR)

Currency Agreement

0.23

0.42

0/1 variable; 1 if currencies of i and j are legal tender in both markets in year t, 0 otherwise (LT)

Customs Liberalization

0.30

0.46

0/1 variable; 1 if customs are liberalized between markets i and j in year t, 0 otherwise (CU)

City Population in 1800

0.39

0.40

Average of the population of city i and city j in the year 1800; in 100,000 people

Market Potential

13.93

1.47

Log distance-weighted state population in year t

Gulden in 1754

0.13

0.34

0/1 variable, equal to 1 if both markets i and j had Gulden as its currency in the year 1754

Currency Population

7.97

8.19

Log distance-weighted population of states that gave each other legal tender status in year t

Distance to the Coast

1.81

1.67

Minimum of market i and market j's distance to the nearest coast, in 100 kilometers

Zollverein Population

10.29

4.26

Log distance-weighted population of states that belonged to the Zollverein in year t

State Railways

0.68

0.33

Average of share of railway in markets in and j that is run by the government

Late Abolition of Serfdom

0.24

0.43

0/1 variable indicating that one or both markets i and j abolished slavery after the year 1831

Length of Observation

11.40

4.71

Number of observations per market-pair

Summary statistics for the baseline sample of twenty-five years before and after train connection, currency agreement, or customs liberalization Statistics weighted by number of observations of market-pair; Number of observations: 6,990

Table 3: Market Expansion and the Relative Costs versus Benefits of Change (1)

(2)

(3)

(4)

(5)

(6)

Panel A: Second-Stage Results Customs Liberalization

-0.173* (0.072)

-0.128# (0.019) -0.211* (0.092)

Train Connection

-0.219# (0.034) -0.158# (0.051)

Currency Agreement

-0.099# (0.030)

Panel B: First-Stage Results Zollverein population

0.022# (0.006)

Distance to Coast

0.014# (0.004)

Market Potential

0.011* (0.005)

City Population in 1800

0.064 (0.050)

Currency Agreement Population

0.020# (0.008)

Gulden in 1754

0.998# (0.004)

F-statistic p-value

< 0.001

0.01

< 0.001

< 0.001

< 0.001

< 0.001

Shea Partial R-squared (percent)

1.7

1.1

2.9

11.4

3.6

9.1

Number of excluded instruments

2

2

2

36

24

39

Panel C: Ordinary Least Squares Customs Liberalization

-0.014 (0.010) -0.025# (0.009)

Train Connection Currency Agreement

Endogeneity test p-value Hansen OverID test p-value

-0.002 (0.015) < 0.01

0.03

< 0.01

0.13

0.37

0.26

Dependent variable: absolute value of percentage bilateral price difference; robust standard errors clustered at the state-pair level in parentheses. All regressions include year- and state-pair fixed effects. #/*/+ Estimate is significant at the 1%/5%/10% level Sample observations are within a twenty-five year span before and after the establishment of train connection, customs liberalization, or currency agreement; number of observations: 6,990; p-value of F-statistics of all regressions < 0.001

Table 4: Institutional and Technological Change Compared (1)

(2)

(3)

Panel A: Second-Stage Results #

Train Connection

-0.129 (0.010)

Customs Liberalization

-0.072# (0.010)

#

-0.162 (0.025)

#

-0.131 (0.008)

#

-0.059 (0.022)

Currency Agreement

-0.031# (0.004)

Institutions

Panel B: First-Stage Statistics Trains First-Stage F-statistic p-value

< 0.001

< 0.001

< 0.001

5.6

4.8

6.1

< 0.001

< 0.001

< 0.001

18.9

9.8

19.4

74

62

96

Hansen OverID test p-value

0.14

0.28

0.16

Number of observations

6,990

6,990

6,990

Shea Partial R-squared (percent)

Insitutions First-Stage F-statistic p-value Shea Partial R-squared (percent)

Number of excluded instruments

Dependent variable: absolute value of percentage bilateral price difference; robust standard errors clustered at the state-pair level in parentheses. All regressions include year- and state-pair fixed effects. #/*/+ Estimate is significant at the 1%/5%/10% level Sample observations are within a twenty-five year span before and after the establishment of a train connection, customs liberalization, or currency agreement; p-value of F-statistics of all regressions < 0.001 Instruments: Population in 1800, market potential (specification (1)-(3)), distance to the coast, Zollverein population (specificaitions (1) and (3)), and Gulden in 1754, currency agreement population (in (2) and (3)), all varying by state

Table 5: Robustness Analysis Baseline (1)

Years 1820 to 1880 (2)

Trimmed 95% sample

(3)

(4)

(5)

(6)

-0.113# (0.011)

-0.129# (0.022)

-0.123# (0.009)

-0.177# (0.022)

Non-ZV customs liberalizations Pervasive Lib'n Only ZV (7) (8)

Panel A: Second-Stage Results Train connection

-0.129# (0.010)

Customs liberalization

-0.072# (0.010)

-0.162# (0.025)

-0.086# (0.011) -0.059# (0.022)

Currency agreement

-0.054# (0.007) -0.091* (0.037)

-0.142# (0.011)

-0.136# (0.012)

-0.053# (0.008)

-0.062# (0.008)

-0.043# (0.016)

Panel B: First-Stage Summary Shea R-squared (%) Trains

5.6

4.8

5.3

4.8

5.6

4.9

5.6

5.6

Institutions

18.9

9.8

17.7

7.9

18.9

10.2

15.5

12.6

74

62

74

62

74

62

74

74

Hansen OverID test p-value

0.14

0.28

0.21

0.32

0.18

0.28

0.13

0.15

Number of observations

6,990

6,990

6,072

6,072

6,642

6,642

6,990

6,990

Number of excluded instruments

Dependent variable: absolute value of percentage bilateral price difference; robust standard errors clustered at the state-pair level in parentheses. #/*/+ Estimate is significant at the 1%/5%/10% level; all regressions include year fixed effects and state-pair fixed effects Sample observations are within a twenty-five year span before and after the establishment of a train connection, a customs liberalization, or a currency agreement; instruments as in Table 4; p-value of F-statistics of all regressions < 0.001

Table 6: The Direct and Indirect Impact of Institutions State Railways (1) (2)

Abolition of Serfdom (3) (4)

Panel A: Second-Stage Results Train connection

-0.222# (0.059)

-0.231# (0.052)

Train * State Railways

0.187* (0.096)

0.125+ (0.067)

Train * Late Abolition of Serfdom

-0.084# (0.032)

Customs liberalization

-0.138# (0.010)

-0.212# (0.021)

0.045# (0.009)

0.026+ (0.015)

-0.067# (0.010) -0.066# (0.017)

Currency agreement

-0.044# (0.013)

State Railways

-0.019* (0.008)

-0.039# (0.013)

Late Abolition of Serfdom

-0.143# (0.006)

0.036# (0.012)

5.9

5.2

11.3

16.4

Panel B: First-Stage Summary Shea R-squared (%) Trains

4.3

4.7

Trains * State Railways

2.4

3.5

Trains * Late Abolition of Serfdom Customs lib'n or currency agr't

6.0

5.3

18.8

10.7

Number of excluded instruments

37

26

74

63

0.11

0.28

0.16

0.15

Hansen OverID test p-value

Dependent variable: absolute value of percentage bilateral price difference; robust standard errors clustered at the state-pair level in parentheses. All regressions include year- and state-pair fixed effects. P-value of F-test of all regressions < 0.001 #/*/+ Estimate is significant at the 1%/5%/10% level. Number of observations: 6,990. Sample of observations within a 25 year window before and after the establishment of train connection, customs liberalization, or currency agreement Instruments: Population in 1800, market potential (specification (1)-(4)), distance to the coast, Zollverein population (specificaitions (1) and (3)), and Gulden in 1754, currency agreement population (in (2) and (4)) P-value of F-test of excluded instruments

Suggest Documents